st: Question on xtivreg

[Subramaniam Ramanarayanan <s-ramanarayanan@kellogg.northwestern.edu>]

Hi,

I am working on a dataset where the unit of observation is a patient in a 
particular year. For each patient, I observe the physician operating on the 
patient, and the hospital the patient is treated in. The data does not 
contain information that identifies patients - I only know the identity of 
the physician and the hospital. In the example below, doc 1 treats 2 
patients in hospital 1 in 1999 and one patient in hospital 3 in 2000.

Row     Physician       Hospital        Year
------- --------------- ------------    --------
1       Doctor 1        Hospital 1      1999
2       Doctor 1        Hospital 1      1999
3       Doctor 2        Hospital 1      1999
4       Doctor 1        Hospital 3      2000
5       Doctor 3        Hospital 2      2000

I would like to run an instrumental variables OLS regression that includes 
Fixed Effects for physician, hospital and year. Are the following two 
specifications equivalent?

Spec 1:         ivreg depvar (endog var = instrument) yeardummies 
hospitaldummies physiciandummies

Spec 2: xtivreg depvar (endog var = instrument) yeardummies 
hospitaldummies, i(physician id) fe

I am not sure if I can use physician id as the "i" variable because the 
dataset is at the level of a patient, and not at the level of a physician.

I am getting different estimates from these two specifications...and am not 
sure which is the correct one to use. Any help will be hugely appreciated.

Thanks a lot!
subbu







____________________________________________________________
Subramaniam Ramanarayanan
Ph.D. Candidate
Managerial Economics & Strategy
Kellogg School of Management
Northwestern University

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/